History of Standard Reference Temperature

1. Introduction

While most dimensional metrologists know that thereference temperature for dimensional measurements is20 °C,1 very few know how or why that temperaturewas chosen. Many people have thought it was, in somesense, arbitrary. In actuality, the decision was the resultof 20 years of thought, discussion, and negotiations thatresulted in the International Committee for Weights and Measures (CIPM)2 unanimous adoption of 20 °C as thereference temperature on April 15, 1931.

In cleaning up my office I found a pack of correspon-dence and documents that shed considerable light onhow this decision was made by the Bureau ofStandards, propagated through U.S. industry, and final-ly brought to closure at CIPM. It is an interesting pieceof history that shows the high level of technical sophis-tication of metrologists of 100 years ago, as well as thedeep commitment of the Bureau’s first Director,Dr. S. W. Stratton, to internationally based standardsof metrology [1].

Volume 112, Number 1, January-February 2007Journal of Research of the National Institute of Standards and Technology

1

[J. Res. Natl. Inst. Stand. Technol. 112, 1-23 (2007)]

20 °C—A Short History of the StandardReference Temperature for Industrial

Dimensional Measurements

Volume 112 Number 1 January-February 2007

Ted Doiron

Precision Engineering DivisionNational Institute of Standardsand Technology,100 Bureau Drive,Gaithersburg, MD 20899-8211

[email protected]

One of the basic principles of dimensionalmetrology is that a part dimension changeswith temperature because of thermalexpansion. Since 1931 industrial lengthshave been defined as the size at 20 °C.This paper discusses the variety ofstandard temperatures that were in usebefore that date, the efforts of C.E.Johansson to meet these variations, andthe effort by the National Bureau ofStandards to bring the United States to theeventual world standard.

Key words: dimensional metrology;gage blocks; gauge blocks; referencetemperature; C. E. Johansson; S. W.Stratton.

Accepted: November 11, 2006

Available online: http://www.nist.gov/jres

Invited Review Article

1 The original manuscripts have a variety of ways of writing thetemperature units. I have changed them to the modern acceptednotation for clarity.2 The General Conference on Weights and Measures (CGPM) is theinternational body made up of official delegates from the countriesthat are signatories of the Convention of the Meter. The International

Committee for Weights and Measures (CIPM) is a group of eminentmetrologists that brings technical matters to the CGPM and is incharge of implementing the CGPM decisions. The InternationalBureau of Weights and Measure (BIPM), in Paris, holds the Inter-national Standards and coordinates the standardization of measure-ments at the various national measurement institutes.

2. The Meter

The U.S. participation in the metric system beganmuch earlier than most people realize, being one of theoriginal signatories of the Convention of the Metre in1875 which set up the International Committees and theInternational Bureau of Weights and Measures in Paris.The U.S. was also an early advocate of expanding therole of the Convention beyond its original scope of massand length to cover a much wider range of measure-ments. The amendment of the original Convention wasratified in 1921, and since that time CIPM has becomethe primary authority for nearly all physical units.

The original Convention chose to base length on theFrench meter bar, which was an end standard. This endstandard was replaced with a line scale, and then furtherreplaced in 1889 with the Platinum-Iridium Meter of theArchives that was the standard until replaced with thewavelength of light in 1960. Thus, until 1960 the meterwas defined as the distance between two lines on theMeter of the Archives at the temperature of the meltingpoint of ice when supported by two cylinders of diame-ter at least 1 cm placed symmetrically under the bar and571 mm apart.

The choice of the melting point of ice seems, today,like a very awkward choice for the reference tempera-ture. It is a difficult temperature at which to actuallymake measurements, both because of human comfortand the fact that the dew point is generally above thistemperature, and some sort of humidity control would beneeded. Temperature was, and still is, a major complicat-ing factor in length measurement. Because all materialschange size with changing temperature, in order todefine the length “1 meter” with a physical object wemust also set the temperature for which it is “1 meter”long.

The standard way to compare length standards was tohave two microscopes, one focused on each line of thesegment of the length scale of interest. Using the micro-scopes, two lengths could be compared to very high pre-cision. The accuracy of the measurement uncertaintydepends on the uncertainty in the temperature of the barduring the measurement. Thermometry in the late 19thcentury was, unfortunately, not nearly as accurate as wasneeded to support the level of accuracy of length com-parisons. The best characterized temperatures at the timewere the melting point of ice and boiling point of water.The melting point was not only the more practical of thetwo; the melting point temperature is 3700 times lessdependent on the atmospheric pressure, and thus muchmore reproducible. The choice of the melting point of ice

was the only acceptable choice for the standard temper-ature for the meter at the time.

3. Reference Temperature for IndustrialMeasurements

This choice did, however, present a problem for indus-trial measurements which are almost never made at sucha low temperature. The basic problem is that if the stan-dard reference temperature is 0 °C, two mating parts ofdifferent materials, say steel and brass, will actually betheir nominal size at 0 °C. If the parts are, instead,assembled at 20 °C, the parts will grow by their coeffi-cients of thermal expansion (CTEs) times the tempera-ture difference from the reference temperature. Since theCTE of steel is about 12 × 10–6/°C and brass is about24 × 10–6/°C, the brass part will expand much more thanthe steel part and assembly may not be possible. Sinceparts are assembled, on average, at around room temper-ature it would seem practical to have the part lengthmeasurement refer to the assembly temperature as close-ly as possible.

The obvious way to overcome this problem would beto take some temperature near room temperature, say20 °C, as the reference temperature for industrial meas-urements and then make a reference bar which wouldhave the same length at 20 °C as the InternationalPrototype Meter has at 0 °C. To accomplish this we needto know the thermal expansion of either the InternationalPrototype Meter or the new standard between 0 °C and20 °C. To use the International Proto-type Meter was nota reasonable choice because to preserve its length it wasused as little as possible, as shown in the BIPM responseto Dr. Stratton of the Bureau of Standards. Dr. Strattonwrote to ask if the U.S. could send its meter to BIPMfor comparison to the International Prototype Meter.Dr. Guillaume, Director of BIPM wrote back (Bureautranslation) [2]:

But comparisons cannot be made with Inter-national Prototype. The International PrototypeMeter as well as the kilogram, and their certifi-cates, are shut up in a depository, which is underthe charge of the Inter-national Committee, andclosed by three locks, one key of which is in myhands, the second is deposited in the Archives ofFrance, and the third is in possession of thePresident of the Committee, Prof. Foerster atBerlin. The depository which is a deep cave underour laboratory, is inaccessible to me as well as toall the world. It cannot be opened and much more

2


3

the prototype can not be taken out except by adecision of the Committee in session.

Thus, the CTE studies had to be made on the replicameters made for routine work at BIPM and for themembers of the Commission. These studies were per-formed, off and on, for the next 30 years at BIPM aswell as other countries National Metrology Institutes(NMIs). As late as the 7th International Conference in1927 there were still major publications on this subject.The work was slow for both technical and bureaucraticreasons. The thermal expansion coefficient was neededto very high accuracy from the melting point of waterup to 20 °C, which is a difficult task made more diffi-cult because at a time the temperature scale itself wasunder serious study. Also, since the InternationalPrototype is not generally available, the studies focusedon the 29 nearly identical copies that were made at thesame time out of the same batch of the Platinum-Iridium alloy. These copies were distributed to themembers of the Convention of the Meter; meter bar No.27 was the legal standard for length in the United Statesuntil the redefinition in terms of the wavelength of lightin 1960. Simply getting the bars back to BIPM formeasurements was time consuming. Efforts began in1921 and continued for 15 years [3].

From a modern perspective, the basic question is thatof uncertainty. Since the meter was defined as the dis-tance between two lines of the International Prototypewhen held at 0 °C, a second meter bar, even if com-pared directly to the International Prototype at 0 °C willhave a larger uncertainty when used at 20 °C: larger bythe uncertainty of the change in the length of the barwhen heated. The size of the uncertainty depends on theknowledge of the CTE of the gage and the accuracy ofthe temperature measurement. For ordinary gagesmeasured in a lab environment, the knowledge of theCTE of the gage and the accuracy of the thermometerswould add considerable uncertainty to the gage at20 °C.

However, for one specific gage these uncertaintiescould be minimized. In particular, at a NationalMeasurement Institute the uncertainty could be reducedto a level that would be negligible for industrial meas-urements. We will see later that C. E. Johansson, inven-tor of the gage block, had done exactly this for his ownuse in his factory. Unfortunately, in the early years ofthe 20th century the concept of uncertainty was still inits infancy, and most of the members of CGPM andscientists of the BIPM seem to have been swayed moreby the philosophical implications than the practicalmerits of 20 °C.

The lines were drawn on this question fairly early, atleast in the United States. In the following exchange ofletters Dr. Guillaume, Director of BIPM, inquires abouta number of standards issues related to the work of theBureau of Standards [4].

I had the pleasure on June 22 of this year towrite you on the following subjects, (1) hydro-meters, (2) thermometers, (3) end standards,(4) question of the carat, (5) units of cold,(6) China, (7) invar tubes.

My letter remaining without reply, I fear that ithas not been received, and I repeat the essentialpoints.…

3. For the end standards, it had to do espe-cially with the reference temperature of 0 °C andto see how it might be received by the Americanindustries.

The reply from Dr. Stratton shows that the questionwas, at least to him, already settled [5].

3. We find in this country a strong objection,even among scientific men, to the use of thetemperature 0 °C as the standard for practicalpurposes, and we are of the opinion that it wouldretard the use of the metric system in this countryto insist upon all measuring apparatus beingcorrect at this low temperature. It seems to methat it would be far better to have the standardsboth of length and capacity agree or nearly agreeat the temperature at which they are commonlyused, say 20 °C, than to have them agree at 0 °C,at which temperature they are rarely used.

For example, to have brass, platinum, steel,and other length measures with widely differentexpansions correct at 0 °C instead of at ordinarytemperatures would not appeal to most users ofsuch measures. The tapes used by our Coast andGeodetic Survey are all made so as to be approx-imately correct at 20 °C. The same is true of allpolariscopic apparatus and of the volumetricapparatus used by chemists. We are strongly ofthe opinion that the temperature of 20 °C shouldbe selected as the temperature at which all work-ing apparatus should be standard.

Despite the U.S. stance, the equally strong stance ofGreat Britain for 62 °F, and the fact that almost no oneoutside of France used 0 °C as the standard reference


4

temperature, in 1915 the 5th International Conferencepassed the following resolution [6].

DECLARATION RELATIVE TO END MEASURES.

Translation.

Considering the precision of adjustment hasbecome an indispensable factor, both for theproper functioning of machines and for manufac-turing in series, an essential element in all indus-trial construction,

Considering the necessity of referring to asingle scale the materialization of dimensionsexpressed numerically by the plan carried out,

Considering that this condition cannot be real-ized except by the choice of a single temperatureof adjustment at which industrial standards shallactually represent their numerical value,

Considering, the International Committee, inits Session of 1909, had already sanctioned foradjustments the temperature of melting ice by theadoption of which the joint action of severalofficial bureaus had already begun the desiredunification.

The Conference declares:

1. Approval of the fixing of the temperature ofmelting ice as the adjustment temperature atwhich end standards intended for control ofindustrial manufacturing should possess theirnominal value.

2. To invite the International Committee toundertake all the necessary work which willassure the highest perfection in construction,determination and use of end standards.

This ironic Declaration seems to have little effect.As an example, in 1915 Dr. Stratton received a letterfrom Dr. R. T. Glazebrook of the National PhysicalLaboratory in Great Britain along with a very detailedanalysis by M. J. E. Sears, Jr., the Deputy Warden ofStandards in Great Britain. In this work Mr. Searsdetails the problems with the 0 °C reference tempera-ture for industrial measurements and makes the case forthe British standard temperature of 62 °F (16.67 °C).

Answering for Dr. Stratton was E. A. Rosa3 [7]:

We have found some discrepancies in theapparent temperatures of standardization of

other metric apparatus o the part of the manufac-turers, due largely, we believe, to lack of appreci-ation of the importance of care in the matter, orto lack of an understanding of the whole subject.We shall be pleased to gather more exact dataas to the practice in this country with a view torecommending a definite policy and shall com-municate the results to you in a later letter.

Unfortunately, the follow-up never was written andDr. Glazebrook wrote back the next year [8]:

On June 3rd, 1915 you wrote acknowledging aletter of ours as to the standard temperature forcommercial metric standards.

You stated you would write further on thereturn of Mr. L.A. Fischer, Chief of the Weightsand Measures Division but no further letter hasreadied us. The matter is assuming increasedimportance and I should value your viewsgreatly. Dr. Guillaume, of course, adheres toStandardization at zero centigrade.

Yours very truly,

R. T. Glazebrook.

At the bottom of the letter is a handwritten note fromDr. Stratton (SWS) Fig. 1.:

While there is no record of the follow up letter, maildelivery was apparently a problem. A large number ofcorrespondences in the files are simply letters acknowl-edging the reception of correspondence, a habit thatwould not be needed if mail delivery was assured. Butit is also true that the number of requests for informa-tion sent to the Bureau of Standards about the standardtemperature problem was rising, and the number ofBureau staff was still quite small.

Dr. Stratton’s answer on October 16, 1916 reveals nochange in his attitude about the problem [9]:

Referring to your letter of August 17, 1916, inregard to a suitable standardization temperaturefor commercial metric standards of length, I haveto say that we have carefully read Mr. Sears’


3 Dr. Edward B. Rosa was hired by Dr. Stratton to head up the elec-trical work when the Bureau was formed in 1901. He died at his deskat the Bureau in 1921.

Fig 1. Hand written note from Dr. Stratton at bottom of letter fromDr. Glazebrook of the National Physical Laboratory (Great Britain).

memorandum, and while we agree with him thatcommercial standards of length, whether metricor Eng1ish, should be standardized at the tem-perature at which they are to be used, we do notconcur in his opinion that the mean work-shoptemperature to be se1ected should be 62 °F.

There is, at the present time, a decided tenden-cy away from the Fahrenheit temperature scale,and we feel that the tendency should be encour-aged. There is, in fact, a bill now pending inCongress by which it is hoped to abolish theFahrenheit scale, at least from Governmentpublications.

The temperature 20 °C is coming more andmore to be accepted as the standard temperaturefor industrial as well as scientific operations. Thesugar industry, for example, is practically on the20 °C basis. All polariscopic tubes, flasks, etc.used in making up sugar solutions are madestandard at that temperature. Very many hydro-meters are standard at this temperature and theglass volumetric apparatus standardized by thisbureau is on that basis and has been for the pastten years or sore. Also many of the steel tapesused in this country are standard at 20 °C.

I might add many other examples to show that20 °C is being largely accepted as the standardtemperature in scientific and technical work.Would it not, therefore, under the circumstances,be better to standardize both the English andmetric commercial standards on this basis ratherthan that of 62 °F? 20 °C would certainly have avery great advantage over 62 °F if urged forinternational adoption; and from. a practicalpoint of view it would be no more difficult tochange the English commercial standard from62 °F to 20 °C (68 °F), than to change the metricstandards from 0 °C to 16.67 °C (62 °F).

Although Dr. Stratton strongly advocated 20 °C asthe reference temperature, the Bureau of Standards,itself, was not completely standardized in its measure-ments. In response to a request from one of theStandards Committees of the American Society ofMechanical Engineers (ASME) the Director asked PaulAgnew4 to review all of the “Standard Temperatures” in

use at the Bureau. His response to ASME, written inJanuary of 1918 shows the span of the problem. Afterdiscussing how 20 °C, 25 °C, and 62 °C were all in usein different fields at the Bureau, the situation world-wide was even worse. He concludes with [10]:

The average laboratory temperatures inEurope are decidedly lower than in nearly thewhole of the United States, and it is doubtfulwhether international agreement is possible onso high a temperature as 25 °C. There has beenmuch difficulty in getting Europeans to agreeeven to 20 °C. Various temperatures have beensuggested and used abroad, 15 °C, 16°C,17.5 °C, 18 °C, etc., for different standardizingpurposes. 15 °C was formerly the one in greatestuse.

I find considerable difference of opinionamong the men at the Bureau. Most of thechemists feel strongly that 20 °C is the tempera-ture that should be used as a standard. Most ofthe electrical men, but not all, prefer 25 °C whilethose in length and volumetric measurementsprefer 20 °C.

In maintaining a temperature for experimentalwork 25 °C is frequently much more convenientthan 20 °, since it is so much easier to heat thanto refrigerate. In electrical work another consid-eration enters. Even if one does go to the troubleto refrigerate in summer, it will frequently happenthat the humidity then becomes very high, even tosaturation, and insulation problems becomeacute.

It seems to be doubtful whether a singlestandard temperature of reference is possible forall the interests involved.

For industrial gages, other than tapes5, however, thequestion had been quietly settled for the U.S. during theFirst World War. The story of the change, basically byfiat of the Bureau of Standards, is summarized byLewis Fisher6 in a letter to Paul Agnew, who was the


5

4 Paul Agnew joined the Electrical Section of the Bureau ofStandards in 1906, and became the Executive Secretary of theAmerican Engineering Standards Committee (forerunner to ANSI) in1919. He served the AESC for almost 30 years. Some of his lettersare as a Bureau employee, but some are on ASME letterhead wherehe was functioning as the Secretary to standards organizations.

5 Surveying tapes that were marked in inches were calibrated to astandard temperature of 62 °F, while those in metric units at 20 °C.When the inch was changed to be 0.0254 meters in 1959, the oldsurveyor inch (1 meter = 39.37 inches) was kept for surveying. Bothpractices continued far after the issue was settled for all other lengthmeasurements.6 Lewis Fisher had worked in the Coast and Geodetic Survey from1880 to the founding of the Bureau, at which time he became one ofthe first members of the new organization as head of Weights andMeasures until his death in 1921.

Executive Secretary of the ASME and the AmericanEngineering Standards Committee (precursor to thecurrent American National Standards Institute (ANSI))[11].

1. Referring to your communication of May27th, subject—Reference Temperature for Steelgages—I have to state that the practice in thiscountry is somewhat mixed, but so far as the gagework is concerned, all gages are made standardat the temperature of 68 °F or 20 °C. This wasdone at the beginning of the war and consequent-ly all gages used during the war were standard-ized at this temperature.

2. The Johansson gages which were sold to usat about the time we adopted this temperaturewere standard at 66 °F. The origin of this temper-ature is entirely unknown to me. It seems to havebeen original to Mr. Johansson, never to myknowledge having been used by anyone else.

The mystery of these 66 °F blocks would not besolved for another few years.

While the Bureau was using 20 °C for nearly alllength other than surveying tapes, it is not clear howwell this information was disseminated to industry. Asexamples, the following two letters came to the Bureaufrom the W. & L. E. Gurley Company and the Brownand Sharpe Company, both manufacturers of precisiondimensional measurement equipment sent these lettersto the Bureau in 1922 [12, 13]:

Gentlemen:Referring to your letter of the 13th instant, rela-tive to the above subject, we wish to call yourattention to the fact that while we handle Tapesonly on a resale basis, we do make YardMeasures, Leveling Rods, Stadia Rods, etc.,which are standard at 62 degrees F.

We are further under the impression that thestandard for weights and measures is 62 degreesF. We are raising this point wondering whetherthere will be any confusion due to one standardfor Tapes and another standard for the line inwhich we are interested, as noted above.

Yours Very Truly,

C. I. Day, General ManagerW. & L. E. Gurley

My dear Mr. Bearce7:

In discussing the question of temperatures atwhich gages should be measured, and in view ofwhat we understand to be the present practice, ofthe Bureau of Standards, using 68 degrees F. asthe working temperature, it would be of interestto us to know how long the Bureau has worked onthe basis of this temperature, and what the rea-sons were for changing from 62 degrees, whichwe understand was used at one time.

Information along these lines would be appre-ciated by

Yours Truly,

Brown & Sharpe Mfg. Co.

The answer from the Bureau was consistently the same[14]:

In reply to your letter of the 13th instant, wewould say that at the time of the entrance of theUnited States into the War, the OrdnanceDepartment proposed to adopt 66 °F as thestandard temperature for gages because at thattime 66 °F was the standard temperature forJohansson gage blocks which were to be the mas-ter reference gages for all gage work. The Bureaucontended that if any change was to be madefrom 62 °F, it should be made to 68 °F (20 °C)which is the standard temperature for the major-ity of physical constants and is the usual tem-perature for laboratory or inspection room work.Furthermore, European countries at the timewere in a transition period from a standard tem-perature of 0 °C for length standards to 20 °C,and it was felt that an international standard tem-perature for length work was very desirable. Wewould say that at the present time both Johanssongage blocks and Pratt & Whitney gage blocks areregularly furnished in this country standard at68 °F. 68 °F has also been adopted by theNational Screw Thread Commission and by theA.S.M.E Sectional Committee on Plain Limitgages, as the standard temperature for gages andgaged products.


6

7 Henry Bearce rose from an assistant in the Gage Section of Weightsand Measures to Director of Weights and Measures over a career thatspanned 1908-1945.

Fortunately, the difference between the inch definedat 62 °F and 68 °F was not a serious problem given themanufacturing tolerances of the time.

4. Expansion of the Scope ofInternational Standards Work

The question of the Standard Reference Temperaturewas dormant for a few years after the war because oflarger issues involving new fields of measurement.Even before the war, the rise in importance of electric-ity to industry was obvious, and the need for standard-ization of the units and measurements grew greatlyduring the war years. Dr. Stratton of the Bureau ofStandards had greatly expanded the electrical capabili-ties during the period, and as the war ended began acampaign to enlarge the scope of the CGPM from justlength and mass to all of the new fields that affectedindustry. The formal resolution, below, was passed atthe March 18, 1919 interim meeting of the NationalResearch Council [15].

On behalf of the Division of Physical Sciences,Leuschner reported on the desirability of enlarg-ing the functions of the International Committeeon Weights and Measures and moved: That theDivision be authorized to appoint a special com-mittee to prepare suitable recommendations inthis respect.

At the first meeting of this new committee, the min-utes show extensive discussion of the subject, and thedecision that, since the matter was very complicated, asmaller subcommittee would be appointed to preparerecommendations for the full committee to consider.This smaller committee was made up of Dr. Joseph S.Ames of Johns Hopkins University, Professor A. A.Michelson of the University of Chicago, Dr. JohnA. Anderson of the Mount Wilson Observatory, andDr. Stratton, Director of the Bureau of Standards.

This proposal was then taken through the CommerceDepartment, led by Herbert Hoover, to the StateDepartment. Since the CGPM is a diplomatic organiza-tion made up of states, the U.S. representative is to beappointed by the State Department. Dr. Stratton was

granted full power to sign the Convention as modifiedat the CGPM meeting of October 27, 1921. A largenumber of changes were made during this meeting ofthe CGPM, including the extension of the scope toelectrical measurements.

After this, the efforts concerning international stan-dardization decreased markedly. Part of the reason waseconomic. During the war the duties and staff of theBureau had grown explosively, from a budget of lessthan $1 million and a staff of 400 in 1914, to over$3 million and 1150 staff in 1918. In the followingyears the decrease was nearly as fast, shrinking to 2/3of the war year levels in 1925. Because of numerouseconomic factors, including a depression in 1920, thisperiod was one of retrenchment and reorganization.Another reason was the departure of Dr. Strattonin January of 1923 to become President of M. I. T.Dr. Stratton had always had a keen interest in standard-ization, and was very active in all phases of bothnational and international standards efforts. On hisleaving, Secretary Hoover appointed him to the VisitingCommittee, a liaison group of prominent men (and laterwomen) from science and industry who were to keepthe Secretary of the Treasury8 apprised of national mat-ters that were in the Bureau of Standards domain, andto report yearly on the work of the Bureau.

During this period, up to 1927, there was little doneon the question of the reference temperature for dimen-sional measurements. The U.S. had been standardizedat 20 °C by the Bureau of Standards, and since theCGPM Declaration of 1915 was widely ignored theissue was still a problem internationally.

5. Proposals to the 7th Meeting of theCGPM in 1927

In early 1927 the Bureau of Standards began to for-mulate a number of proposals to the 7th meeting of theCGPM, which was to be held in October 1927. Theprocess ended up with five [16]:


7

8 The original Bureau of Standards was part of the Department of theTreasury. When the Department of Labor and Commerce was formedin 1903 the Bureau was moved into the new department.

1. Recommend that the Conference adopt the wave-length of the red radiation of Cadmium vapor as thefundamental standard for the wavelength of light, andindirectly for the meter. A committee was formed tolook into the matter. (Prof. Kösters of Germany hadfound that the light from Krypton gas had better prop-erties than Cadmium, and the group decided to makefurther studies of the properties of Krypton lamps.Krypton lamps were made the standard for length in1960.)

2. Recommend that the Conference adopt the rela-tion 1 inch equals 0.0254 meters. This was consideredout of the scope of the international body since it didnot involve the metric system. It was referred to theparties involved (English speaking countries) for reso-lution. (The value 1 in = 25.4 mm was adopted by theNational Measurement Institutes of the countries thatstill used inches in 1959).

3. Recommend that the Conference adopt 20 °C(68 °F) as the standard temperature at which industrialstandards of length shall have their correct nominallength. A committee of five members was set up tostudy this question and report before the first of March,1929. The following were the members:

Bureau of Standards, Washington;Laboratoir d’Essais du Conservatoire des Arts

et Metiers, Paris;National Physical Laboratory, Teddington;Physikalisch-Technische Reichsanstalt,

Charlottenburg;Director Guillaume of the International Bureau.

4. Recommend that the Conference consider thedesirability of having all certificates issued by theInternational Bureau of Weights and Measures containall relevant test data to permit conditions of comparisonto be known and reproduced, in order that reduction tosuch other standard conditions as may be required byNational Standardizing Laboratories may be made withthe highest accuracy.

5. In view of the ever increasing demands for aknowledge of the exact lengths of the nationalPrototype Meter Standards, it is recommended thatthe Conference urge upon the International Bureau of

Weights and Measures the desirability of expediting thestudy of the lengths and coefficients of thermal expan-sion of these standards, in order that any necessarychanges of certificates may be authorized by the EighthGeneral Conference.

6. A U.S. Industry Consensus

Work on all of these, except the issue of the inch,were taken up very quickly. The Director of the Bureauof Standards, Dr. Burgess who succeeded Dr. Strattonin 1923, immediately began a round of correspondencewith interested parties in the United States to assess theimpact of the change in reference temperature. HenryBearce, head of the Weights and Measures Section ofthe Bureau. prepared a list of important industry andgovernment representatives to invite to a conference todecide the proper course for the U.S. delegates to theConference [17]. Chosen were:

Bureau of Standards, Washington, D.C.Gage Division, Ordnance Dept., U.S.A.,

Washington, D.C.Engineering Bureau, Navy Department,

Washington, D.C.National Screw Thread Commission, Washington,

D.C.American Society of Mechanical Engineers,

29 West 39th St., New York, N.Y.Society of Automotive Engineers, 29 West 39th St.,

New York, N.Y.National Machine Tool Builders’ Association,

Cincinnati, OhioAmerican Railway Association, 30 Vesey St.,

New York CityPratt & Whitney Co., Hartford, Conn.Brown & Sharpe Mfg. Co., Providence, R. I.Johansson Div., Ford Motor Co., Detroit, Mich.Taft-Peirce Mfg. Co., Woonsocket, R. I.Greenfield Tap & Die Corp., Greenfield, Mass.Sheffield Mach. & Tool Co., Dayton, OhioStandard Gage Co., Poughkeepsie, N. Y.The L. S. Starrett Co., Athol, Mass.The Van Keuren Co., 12 Copeland St., Watertown,

Boston 72, Mass.Lt. Col. E.C. Peck, Room 305 Lake Erie Bank Bldg.,

1612 Euclid Ave. Clev., Ohio.S. W. S. Stratton, Pres. Mass. Inst. of Tech.,

Cambridge, Mass.


8

On April 13, 1928 Dr. Burgess sent identical lettersto all of the companies and individuals on Bearce’s list.The letter said:

At the last meetings of the InternationalCommittee of Weights and Measures and theGeneral Conference on Weights and Measuresheld in Parts last September and October, thequestion of international agreement on the stan-dard temperature for intercomparison of indus-trial standards of length, such as precision gageblocks and other end standards, graduatedscales, lead screws, etc., was discussed, and therewas set up a committee of five members to studythis question and report before the first of March,1929. The following are the members: Bureau ofStandards, Washington; Laboratoir d’Essais duConservatoire des Arts et Metiers, Paris; NationalPhysical Laboratory, Teddington; Physikalisch-Technische Reichsanstalt, Charlottenburg; withDirector Guillaume of the International Bureau.

The American delegates supported by practi-ca1ly all the Europeans except the British andFrench proposed 20 degrees C (68 degrees F),the British 62 degrees F (16 2/3 degrees C),andin France I believe both 0 degrees C (32 degreesF) and 20 degrees C (68 degrees F) are in use.

It is considered desirable to get a consensus ofopinion from the engineering and industrialactivities of the various countries and with thisobject I am cal1ing a conference at an early dateat the Bureau of Standards and request yourorganization to designate a representative.

The early responses were both unanimous and defi-nite that 20 °C was the proper choice.

“Our practice is invariably to give 68 degreesFahrenheit as standard, and we have no doubtthat this practice is well nigh universal in theengineering and inspection departments ofAmerican Industry.”

1928-04-16 Taft-Pierce response

"Replying to yours of the 13th, we certainly favorthe use of 68 degrees Fahrenheit as a standardtemperature for intercomparison of gauges, etc."

1928-04-17 Brown & Sharpe response

"In reply to your letter of April 13th, we wish tobe placed on record as favoring 68 °F. (20 °C.) asthe standard temperature for intercomparison ofstandards of length. This temperature is onewhich most nearly approaches the average shopcondition, and is a temperature in which produc-tion can be maintained as efficient standards.

1928-04-19 Pratt & Whitney response

Because of the swift and unanimous responses,Dr. Burgess decided that the matter would not need aconference and wrote again to see if there were anyobjections to settling the matter by mail. There were noobjections and all of the responses, but one, were defi-nite on the matter that 20 °C was the preferred refer-ence temperature. The only other response was fromthe Society of Automotive Engineers, which suggested“rather detailed questionnaire asking specifically aboutpresent practice by individual companies, to what theirpractice applies, and other leading questions would bea very helpful index as to whether a general conferenceshould not be called later on. It would also indicatewhat reference temperatures are most generally in use.”By the end of May, the date of this response, thequestion was already settled.

With the U.S. position now set by industry andgovernment preference, the question moved to theinternational arena.

7. The French Proposals

There were two large works discussing the problemfrom the French perspective. One, an article in LeGenie Civil by A. Perard of the BIPM, and a secondlong report by M. F. Cellerier, Director of Laboratoird’Essais du Conservatoir National des Arts et Metiers.Cellerier discusses a number of proposals and favorsone by M. Gaux that is very similar to that of Perard.

The proposal, despite the strong reactions against it,is not as strange as it first appears if compared to massmetrology at the time. In mass metrology the largestenvironmental factor is air buoyancy. From very earlytimes [18], mass metrologists had made allowance forthis factor for industrial measurements using the con-cept of “apparent” or “conventional” mass. The appar-ent mass was the true mass with the correction for airbuoyancy that would be accurate if the weight had thesame density as brass (later revised to 8.0 gm/cc) andthe density of air was 0.0012 gm/cc. For most metal


9

weights, the assumed buoyancy correction was within10 % of the true correction and for industrial purposesthe system was accurate enough that further correctionswere not needed. This system is still in use for mostcommercial weighing systems.

In analogy to “conventional” mass, the Graux [19]and Peraud [20] proposals would have effectivelyinvented a quantity “valeur-type” which might beunderstood as “conventional” length. The largest envi-ronmental factor in length measurement is temperature,so the conventional length would differ from the truelength (referred to the standard temperature 0 °C ) byassuming a common CTE (11 × 10–6 / °C ) and industri-al reference temperature (20 °C) that would be accurateenough for much industrial measurement. Since mostindustrial parts are iron or steel, this method wouldreduce the error from thermal expansion by 90 %,which would be adequate for many industrial measure-ments. Unfortunately, the relative size of thermal errorsin length measurements and mass measurementsare quite different, and while the “conventional” massidea is still useful today, the “conventional” lengthwould have been inadequate for precision parts even in1927.

The suggestion was forcefully rejected at the Bureauby a number of scientists, as in the example below:

The proposal of Perard is impractical for severalreasons. First, it assumes a coefficient of expan-sion and employs it over a range of 20 °C where-as it is known that gage blocks have coefficientsdiffering somewhat from block to block and dif-fering considerably from the proposed assumedvalue.

In the second place an unnecessary complica-tion is introduced. in the computation, a compli-cation which is confusing. A quantity, termed byPerard, “valeur-type”, is introduced the physicalmeaning of which is not readily apparent.1927-09-08 Judson to Burgess on Perard paper

Perard’s suggestions did not fare any better inEurope. In this excerpt from a letter from Dr. Kösters inGermany to Dr. Stratton (translated by Prof. Vogel)dated August 27, 1928 he gives the German position:

1. Adjusting temperature for industry measures.. . .For 1. they are of the opinion in Germany, thatone can no longer depart from 20 °C, since 20 °Calone gives a clear and plain definition, and, onecould no longer return to 0 °C without giving upthe attained precision again. By Perard. a propo-sition for comparison is now made in an essay.The measure with the designation of "1 m,"is said. to be at 20 °C = 1 m + 11 × 20 µ, nomatter of what material the measure consists.This proposition will not be accepted under anycircumstances in Germany 9, since it is whollyunclear, and, departs from the ideal conception ofthe meter as an independent unchangeablelength.

The temperature of 0 °C is also not debatablesince at 0 °C one can not measure. I presupposethat you are of the same views and I figure thatyou cling to 20 °C = 68 °F.

In another report for the 1931 CIPM meeting, M. J.E. Sears, Jr., Deputy Warden of Standards for GreatBritain, presented the case for 62 °F as a standard [21]. One interesting fact brought up by Sears is that thepressure scale, which is measured in inches or mm ofmercury would be changed if the reference temperaturefor length were modified. New tables would be neededfor mercury barometers. The final paragraph, however,reported that Great Britain and NPL would change thenew standard temperature if international agreementwas obtained.

8. C. E. Johansson Letter

One of the most important reports was from the mostwidely known and respected length metrologist in theworld, C. E. Johansson. In 1928 Dr. Burgess, of theBureau, asked for Mr. Johansson's thoughts on thequestion of International agreement on the standardtemperature for industrial length measurements.


10

9 Underlined in the original.

Mr. Johansson responded with a history of his efforts instandardizing length standards, as well as his philosophyof measurement and recommendations. The entire letteris presented as the appendix to this paper. Notice the pen-cil marks which denote changes to be made for transla-tion into French and submission to the CGPM as areport.

The letter displays the great sophistication ofMr. Johansson as a metrologist, and shows that he hadrecognized the problems with the varying reference tem-peratures almost from the beginning of his invention ofgage blocks. There are actually two letters, one was sentto Dr. Burgess and was taken to France for the meeting.Mr. Johansson discovered that he had made a mistake inhis letter and sent a revised copy to the hotel in Pariswhere Dr. Burgess was to be staying (See Fig. 2). Thesection that was revised described his efforts to assign astandard reference temperature, and the specific changewas to add the story of the 66 °F gage blocks sets that theBureau received from the Ordnance Department duringWorld War I.

Because of his continuing interest in internationalstandards, and his standing in the world metrology com-munity, Dr. Stratton had been retained as a U.S. officialdelegate to the CGPM even after he left the Bureau in

1923 and continued until his death in October, 1931. Inwhat was probably his final act as the U.S. represen-tative, Dr. Stratton cast his vote to adopt 20 °Cas the industrial reference temperature in April, 1931[22].

Johansson’s most interesting, and one of the earliestexperiments was in 1903 when he sent a nominal100 mm gage to BIPM and asked for the temperature atwhich the gage was exactly 100 mm long. His eventu-al answer a year later [23] was that the block was, toexperimental uncertainty, exactly 100 mm at 20.63 °C.He then made a block 0.0007 mm longer, one thatwould be an “absolute” measuring standard of 100 mmat 20 °C. To check on his process as it evolved, in 1912he sent four blocks 100 mm, 50 mm, and two 25 mmblocks to BIPM for calibration. At 20 °C, all werefound to be their nominal size to better than 0.1micrometer.

When Johansson began making length standards itwas obvious that different customers had different ref-erence temperatures. His solution was to buy measur-ing equipment and standards from different sources,measure them against his standards that had been meas-ured at BIPM, and assign the reference temperature foreach customer through his experiments. He found, for


11

Fig. 2. Telegram from C. E. Johansson to Dr. Burgess en route to the CGPM meeting in Paris.

example, Brown & Shape Company in the U.S. used62 °F, Ludwig Lowe, Berlin used 25 °C, Reinecker,Chemnitz used 14 °C, and the State Works of France0 °C. Early sets for Japan were made for 62 °F, but after 1926 when the metric system was adopted all sets were adjusted for 20 °C. Also, from his work he found thatthe average temperature in laboratories and shopswhere his gages were used was about 20 °C, so he usedthat as the default unless the customer specified differ-ently.

In another experiment in 1905, he sent a 3 inchstandard to the Bureau of Standards in Washington forcalibration. The Bureau issued a report giving thelength of the standard at 27.2 °C. Knowing the coeffi-cient of expansion of the block and his measurement ofthe block he found that the Bureau would be “nearlycorrect” for a reference temperature of about 66 °F. Hethen made blocks for the American market, primarilythe War Department, that were marked as 66 °F. Thisodd reference temperature was not noted until theOrdnance Department began working with the Bureauof Standards during World War I. Johansson then saysthat he had always made metric blocks to the 20 °Creference temperature, and after his excellent agree-ment with BIPM in 1912 he made all inch size blocksfor the American market to the 20 °C reference temper-ature. It does not appear that anyone in the Statesnoticed the change.

Besides the practical advantage that most countrieswere using 20 °C as the reference temperature,Johansson mentions two other advantages. First, whenhe made gages at 62 °F the effects of the operators'body heat affected the gages faster, reducing theamount of time available for final adjustments oflength. The second was that 68 °F and 20 °C were bothintegers, and calculations were easier.

He finishes with a discussion of his principles formeasuring. For shop inspection he advises [24]:

V Working Temperature

a. Gage Work. The temperature in the work-ing-room, the work pieces and the gages shouldcorrespond with each other, and in accuratework, i.e., manufacturing, adjusting and calibrat-ing of high precision gages and the like, the stan-dard temperature + 20 °C = + 68 °F, must beheld.

b. Production Work. The ideal condition forproducing and measuring of interchangeablemachine parts and the like would be to do allwork and inspection at 20 °C – 68 °F, then themost severe trouble would be eliminated whatmeasuring-values are concerned.

But when this can not be done, the next bestand more possible way is to use cooling plates orother means for bringing the Work and the Gagesto the same temperature. In. this case the differ-ence in coefficient of expansion only wouldchange the measuring-value, and in most cases itwould be found to be well within the given toler-ance and thus a cheap and good interchangeableproduct can be obtained.

This last idea is still the core idea for inspection ofhigh precision parts, an idea that is mentioned in veryfew textbooks on the subject of dimensional metrologyand inspection.

8. International Agreement of StandardReference Temperature

The issue was finalized at the CIPM meeting onApril 15, 1931. With no discussion and a voice vote, thefollowing proposal was adopted unanimously [25]:

As the normal temperature for adjustment ofindustrial standards, the Committee adopts thetemperature of 20 °.

The adoption of 20 °C as the reference temperature forlength measurements was later incorporated into theinternational standards system as the first standardof the International Organization for Standardization(ISO 1) in 1951. In the United States, the NationalStandard ASME/ANSI Y14.5: Dimensioning and Toler-ancing assigned the default temperature of 20 °C for alldimensional drawings.

There have been occasional attempts to change thereference temperature, the latest in the early 1990s [26].These efforts have failed primarily because of the largecost of implementing the change. In 1931 the cost of thechange was restricted to only a few countries and thecosts were manageable; neither of these conditions holdstrue today. Given the very tight tolerances for modernparts and the large effect of temperature, it is unlikelythat the reference temperature can be changed again.


12

9. References[1] All of the notes on Bureau staff are derived from R.C.

Cochrane, Measures for Progress, National Bureau ofStandards, 1966.

[2] 1909-11-13, Letter, Guillaume to Stratton.[3] The International Bureau of Weights and Measures 1875-1975,

translation of the BIPM Centennial Volume, NBS SpecialPublication 420, May 1975.

[4] 1909-10-21, Letter, Guillaume (BIPM) to SWS.[5] 1909-11-13, Letter, Stratton to Guillaume.[6] Bureau of Standards translation, Fifth General Conference of

Weights and Measures, 1915.[7] 1915-06-03, E. B. Rosa to Glazebrook.[8] 1916-08-17, Dr. Glazebrook (NPL) to Stratton.[9] 1916-10-16, Letter, Stratton to Dr. Glazebrook (NPL).

[10] 1918-01-11, Letter, Agnew to F. P. Cox, Manager of the WestLynn plant of General Electric.

[11] 1920-06-18, Fisher to Agnew.[12] 1922-07-21, Letter, Gurley to Bureau of Standards.[13] 1922-04-13, Letter, Brown & Sharpe to Bearce.[14] 1922-04-26, Letter, Bearce to Brown & Sharpe.[15] Minutes of National Research Council meeting of 1919-03-18.[16] 1927-08-08, U.S. Proposals in English with markups.[17] 1928-04-13, Memorandum from Bearce to Burgess.[18] NBS Circular No. 3, Design and Test of Standards of Mass,”

3rd edition, 1918.[19] Rapport de M. F. Cellerier in Procès-Verbaux Des Séances,

Comité International des Poids et Mesures, Session de 1929.[20] A. Peraud, La Température d’Ajustage des Calibres Indsutriels,

Génie Civil du 25 juin, p. 622, 1927.

[21] Rapport de J. J. E. Sears, Jr., Sur la Témperature d’Ajustage desÉtalons Industriels de Longueur, in Procès-Verbaux DesSéances, Comité International des Poids et Mesures, Session de1931, p71.

[22] Procès-Verbaux Des Séances, Comité International des Poids etMesures, Session de 1931.

[23] C. E. Johansson 1864-1943: The Master of Measurement, byTorsten K. W. Althin, published by C. E. Johansson AB,Stockholm, 1948.

[24] 1928-10-25 Letter C. E. Johansson to Stratton[25] Comme température normale d’ajustage des mesures indus-

trielles, le Comité adopte la temperature de 20 °C, from Procès-Verbaux Des Séances, Comité International des Poids etMesures, Session de 1931, p. 63. Translation in report of themeeting prepared by Mr. Crittenden for Dr. Burgess and sent toDr. Stratton for his approval, June 2, 1931.

[26] K. Blaedel and F. Parsons, ISO Studying ReferenceTemperature Change, Quality Magazine, Hitchcock Pub., Feb.,1993.

About the author: Ted Doiron is a physicist and theGroup Leader of the Engineering Metrology Group inthe Precision Engineering Division of NIST. TheNational Institute of Standards and Technology is anagency of the Technology Administration, U.S.Department of Commerce.

10. Appendix

The second letter from C. E. Johansson to Dr. Burgessis presented here in its entirety. The other documentscited in this paper can be found on the EngineeringMetrology Toolbox website,http://emtoolbox.nist.gov/Main/Main.asp.


13


14

15


16


17


18


19


20


21


22


23


History of Standard Reference Temperature

Documents

Transcript of History of Standard Reference Temperature